FLOURY ENDOSPERM19 encoding a class I glutamine amidotransferase affects grain quality in rice

As a staple food for more than half of the world’s population, the importance of rice is self-evident. Compared with ordinary rice, rice cultivars with superior eating quality and appearance quality are more popular with consumers due to their unique taste and ornamental value, even if their price is much higher. Appearance quality and CEQ (cooking and eating quality) are two very important aspects in the evaluation of rice quality. Here, we performed a genome-wide association study on floury endosperm in a diverse panel of 533 cultivated rice accessions. We identified a batch of potential floury genes and prioritize one (LOC_Os03g48060) for functional analyses. Two floury outer endosperm mutants (flo19-1 and flo19-2) were generated through editing LOC_Os03g48060 (named as FLO19 in this study), which encodes a class I glutamine amidotransferase. The different performances of the two mutants in various storage substances directly led to completely different changes in CEQ. The mutation of FLO19 gene caused the damage of carbon and nitrogen metabolism in rice, which affected the normal growth and development of rice, including decreased plant height and yield loss by decreased grain filling rate. Through haplotype analysis, we identified a haplotype of FLO19 that can improve both CEQ and appearance quality of rice, Hap2, which provides a selection target for rice quality improvement, especially for high-yield indica rice varieties.


Introduction
As one of the three staple foods in the world, the yield and quality of rice have always been the focus of scientists and breeders . After the tide of the Green Revolution, rice yield has been greatly improved ), but the development of quality breeding was relatively stunted (Zhang et al. 2020a). The CEQ and appearance quality of rice directly determine its market price and consumer preference. Therefore, it is urgent for breeders to Abstract As a staple food for more than half of the world's population, the importance of rice is selfevident. Compared with ordinary rice, rice cultivars with superior eating quality and appearance quality are more popular with consumers due to their unique taste and ornamental value, even if their price is much higher. Appearance quality and CEQ (cooking and eating quality) are two very important aspects in the evaluation of rice quality. Here, we performed a genome-wide association study on floury endosperm in a diverse panel of 533 cultivated rice accessions. We identified a batch of potential floury genes and prioritize one (LOC_Os03g48060) for functional analyses. Two floury outer endosperm mutants (flo19-1 and flo19-2) were generated through editing improve the quality of rice, whether it is to increase farmers' income or meet consumers' higher demand, and the prerequisite for this is that there are abundant genetic resources available for the improvement of rice quality.
Starch and protein, as the two most important components in endosperm, are also the two most important factors affecting the CEQ (Hori et al. 2016). Amylose (AC), gel consistency (GC), and gelatinization temperature (GT) are three main indicators for physical and chemical properties of the starch in rice endosperm (Sun et al. 2006), which directly affected cooking and eating quality. Since Wx gene was responsible for the synthesis of amylose (AC), and gel consistency (GC) was negatively correlated with amylose (AC), both GC and AC were controlled by Wx (Zhang et al. 2020a). The synthesis of amylopectin is more complex than that ofamylose, which mainly involves soluble starch synthase (SSS), starch branching enzyme (SBE), and starch debranching enzyme (DBE) (Yang et al. 2018a). The gelatinization temperature (GT) was mainly controlled by ALK gene, which encodes SSSII-3 (Gao et al. 2011). The base substitutions in ALK cause amino acid changes in SSSII-3 and result in the alteration of enzymatic activity, and thereafter alter the fine structure of crystalline lamellae of amylopectin, and are finally reflected in GT (Gao et al. 2011). So far, most of the researches on CEQ have focused on the abundant Wx allelic variations and ALK gene (Zhang et al. , 2020bHori et al. 2021;Zhou et al. 2020), or the interaction between them (Yang et al. 2018a). Protein content is also an important factor affecting CEQ. Previous reports have revealed a significantly negative correlation between the protein content and CEQ (Okadome 2005;Martin and Fitzgerald 2002). Although many QTLs affecting protein content have been reported (Yano et al. 2016;Bhattarai and Subudhi 2018;Kashiwagi and Munakata 2018;Zhao et al. 2011), only OsAAP6 (Peng et al. 2014 and OsGluA2 (Yang et al. 2019), both are positive regulators of protein content, have been cloned by forward genetic methods so far. Chalk5 ), a major chalky gene previously reported, is a negative regulator of protein content. These genes regulate protein content and also affect amylose and gel consistency Peng et al. 2014;Yang et al. 2019).
Transglutaminases (protein-glutamine c-glutamyl transferase, EC 2.3.2.13, TG) are a class of enzymes catalyzing an acyl-transfer reaction between the γ-carboxamide group of a peptidebound glutamine and the ε-amino group of a lysine (Büttner et al. 2011). TGs were found to be widely distributed in microorganisms, plants, invertebrates, amphibians, fish, and birds (Beninati and Piacentini 2004). Although microbial transglutaminase has been widely used in food (Kieliszek and Misiewicz 2014), the transglutaminase genes from animals and plants are rarely studied (Beninati and Piacentini 2004), especially in plants, only one gene GAT1_2.1 gene (At1g15040) has been explored in Arabidopsis thaliana (Zhu and Kranz 2012), without any related report on transglutaminase genes in rice.
In this study, we carried out GWAS with a population consisting of 533 rice varieties and further functional analysis of one predicted floury gene FLO19. We showed that FLO19 gene plays an important role in regulating rice quality and maintaining normal carbon and nitrogen metabolism in rice. Haplotype analysis provides a valuable elite haplotype of FLO19 gene for breeders seeking to improve the quality of high-yield indica rice varieties.

Plant materials and growth conditions
A panel of 533 Asian cultivated accessions was employed in the GWAS analysis, which are landraces and elite varieties collected from 87 countries, and consist of 296 Ind accessions, 156 Jap accessions, 49 Aus accessions, and 32 Others (Xie et al. 2015). Information about these accessions is listed in Data S1, including serial number, names, and subpopulation classification. These accessions were grown at the Experimental Station of Huazhong Agricultural University in Wuhan, Hubei, during the 2015 growing season. Two FLO19 mutants and wild-type Zhon-ghua11 (ZH11) were grown at the Experimental Stations of Huazhong Agricultural University, Wuhan (Hubei) and Lingshui (Hainan), during 2017-2019 growing seasons.
Twelve plants for each 533 accession and thirty-six plants for each line of FLO19 were transplanted in a row with 16.5 cm between plants and 26 cm between rows. All materials were grown in a completely randomized design. Field management followed normal agricultural practices.

Vector construction and transformation
The CRISPR/Cas9 system was used to generate knock-out mutants . The target sgRNA GGG CCG ATC GAA GAA GGG CT for FLO19 (LOC_Os03g48060) was inserted into intermediate vectors pER8-Cas9-U6 or pER8-Cas9-U3 and finally cloned them into pCXUN-Cas9. The correct construct confirmed by sequencing was introduced to Agrobacterium tumefaciens EHA105 and transformed into rice cultivar ZH11. A pair of specific InDel primer D5 was used to screen T 0 transgenic-positive plants. DC5-F/DC5-R was applied to amplify the target sites, which were sequenced to validate the mutation in T 0 and further confirmed in the T 1 generation. The relevant primers are listed in Table S1.

Microscopy
For scanning electron microscopy, brown rice grains were transversely broken by two forceps, coated with gold under vacuum conditions, and examined with a scanning electron microscope (JSM-6390LV, JEOL) at an accelerating voltage of 10 kV and a spot size of 30 nm. Scanning electron microscopy analysis was based on at least three biological replications of the mounted specimens. All procedures were carried out according to the manufacturer's protocol.
For transmission electron microscopy, the sections of endosperm at 15 DAF were examined with a transmission electron microscope (H-7650, HITACHI).

Measurement of various quality traits
The determination of four main storage proteins, amylose content, alkali spreading value, gel consistency, and brown rice protein content was performed as previously described Tan et al. 1999;Mariotti et al. 2010;Wu et al. 2020). The corresponding biological and technical repetitions have been indicated in the relevant figures. A rice appearance quality inspection analyzer (SC-E type, Wanshen) was used to evaluate the chalkiness rate and 200-300 milled rice grains per sample were subjected to evaluation. Endosperm at 15 DAF was collected to detect the content of sucrose, glucose, and fructose using a kit purchased from Grace Biotechnology Company (Suzhou). All procedures were carried out according to the manufacturer's instructions. Each sample was measured with three biological replications.
Hydroponic culture assay and enzyme activity assay A standard rice culture solution for hydroponic experiments was used as previously described (Yang et al. 2018b). 0 N, 0.5 N, 1 N, 2 N, and 4 N represent nitrogen-deficient, 0.5-fold nitrogen, onefold nitrogen, twofold nitrogen, and fourfold nitrogen nutrient solution, respectively. The length of shoot and root was determined using ImageJ software (US National Institutes of Health). Plant tissues were immediately frozen in liquid nitrogen after sampling and stored at − 80 ℃ before use. Activities of glutamine synthase, glutamate synthase, asparagine synthetase, and glutamine amidotransferase were determined using kits purchased from Grace Biotechnology Company (Suzhou), and analyzed according to the manufacturer's instructions. Three plants were mixed in one 1 3 replication, and three biological replicates were conducted for each line.

RNA preparation and qRT-PCR
Total RNA was isolated with Trizol reagent (Invitrogen) according to the manufacturer's instructions from flag leaf, root, sheath, stem, young panicle, pulvinus, seedlings, and endosperm at 7 DAF of ZH11 and flo19 mutants. qRT-PCR assay was performed as previously described (Zhou et al. 2020). Each measurement was determined in at least two biological samples and three replicates for each sample. The relevant primers are listed in Table S1.

Haplotype analysis
The SNP and InDel variation data for FLO19 (LOC_ Os03g48060) in all 533 accessions are available at RiceVarMap (http:// ricev armap. ncpgr. cn/). The haplotypes were classified according to SNPs in exons and UTR regions.
Genome-wide association analyses GWAS analysis was performed as previously described (Zhou et al. 2017). Only SNPs with a minor allele frequency (MAF) > 5% and a missing rate < 15% were selected for association analyses. We performed the analysis using the linear mixed model provided by the EMMAX program (Kang et al. 2010). Prior to statistical analyses, phenotypic data was transformed using Warped-LMM software or square root Box-Cox transformation to minimize departure of the data from model assumptions (Gurka et al. 2006;Fusi et al. 2014). The genome-wide significance thresholds of the GWAS were determined using a modified Bonferroni correction as described by Li et al. (2012), in which the total number of SNPs (M) for threshold calculation was replaced by the effective number of SNPs (Me). The physical locations of the SNPs were identified based on the rice annotation version 7.0 of variety Nipponbare from Michigan State University.

Statistical analysis
Histograms, boxplots, and GWAS analyses were constructed using phenotypic grand means. All the bar charts, box charts, pie charts, and line charts in this study were drawn by GraphPad Prism 8. Restricted maximum likelihood estimates of genetic variance were calculated using the mixed.solve() function in the R package rrBLUP (version 4.4) and the value was divided by the total phenotypic variance. The best linear unbiased predictors (BLUPs) for genetic values were calculated using the mixed.solve() function in R package rrBLUPs. Differences between two groups were examined by Student's t-tests. Differences among multiple groups were analyzed through one-way ANOVA and Duncan multiple comparison by using IBM SPSS statistics 16.0 software. The preliminary processing and analysis of phenotypic data were completed by Microsoft Excel 2016.

Artificial creation of the flo19 mutants
In order to reveal the genetic basis of floury endosperm in rice, we evaluated floury endosperm phenotype of 533 Oryza sativa accessions (Fig. 1a, Data S1), which contains worldwide landraces and elite varieties (Xie et al. 2015). The rice accessions with floury endosperm phenotype accounted for only a small part in the whole population, including 22 Ind accessions and 31 Jap accessions (Fig. 1b, Data S1). Subsequently, we performed a genome-wide association study on floury endosperm using 6.5 million SNPs characterized in the whole population, and detected 343 significant loci on 12 chromosomes (Fig. S1, Data S2). The most significant association signal for floury endosperm was observed on chromosome 6, and the lead SNP was located to the promoter of Wx, the major gene for amylose content (Fig. S1, Data S2). However, casual genes underlying most significant loci remained to be characterized (Data S2). Among those, we noticed that LOC_Os03g48060 encoding the class I glutamine amidotransferase was located at 70 kb downstream of the lead SNP on chromosome 3 and presumably participate in the process of nitrogen assimilation to affect the synthesis and accumulation of stored substances (Data S2), while the production of floury endosperm phenotype is often related to the change of stored substances. To further investigate the function of LOC_Os03g48060, we produced two mutants using a CRISPR/Cas9 vector targeting the second exon (Fig. 1c), of which both showed consistent floury endosperm phenotype (Fig. 1d). Therefore, LOC_Os03g48060 was named as FLO19 and the two mutants were marked as flo19-1 and flo19-2, respectively.
Phenotypic characterization of the flo19 mutants Compared with wild ZH11, the endosperm of the two mutants was opaque and white (Fig. 1d). Interestingly, cross-sectional analysis showed that the peripheral region of flo19 mutant grain appeared floury-white, while the inner endosperm was translucent, as in wild-type endosperm (Fig. 1e). Scanning electron microscopy analysis revealed that the outer endosperm cells of flo19 mutant grain were packed with loosely arranged composite starch granules, which was quite different from the dense irregular polyhedral starch granules in wild-type ZH11 (Fig. 1f, g). To better dissect developmental defects in the endosperm of flo19 mutants, sections made from endosperm at 15 DAF (15 days after flowering) were used for transmission electron microscopy analysis. The amyloplasts in peripheral endosperm cells of ZH11 were filled with regular polyhedral starch granules, while the amyloplasts in flo19 mutant showed a fragmented state and starch granules were scattered in the matrix between the amyloplasts (Fig. 1h, i). Both the two mutants showed reduced height (Fig. 2a, b), but no difference on tiller number and grain number per spike, relative to wild ZH11 (Fig. S2a, b). By monitoring the dynamic filling process of ZH11 and mutants, we found that the endosperms of the mutants were flatter than those of ZH11 from 11 DAF, and this difference in fullness became more and more obvious with the passage of time (Fig. 2c). Correspondingly, the filling rate of flo19 mutants was gradually lower than that Fig. 2 Phenotypic analysis of the flo19 mutant. a Comparison of plant architecture between ZH11 and mutants. Scale bars, 10 cm. b Plant height statistics of ZH11 and mutants (n ≥ 41). Different letters represent significant differences at P < 0.05, Duncan's multiple range test. c Endosperm of ZH11 and flo19 mutants at different filling stages. d The accumulation of dry weight of ZH11 and flo19 mutants at dynamic grain filling stage. Asterisks indicate statistical significance compared with the wild type (n = 3). e 1000 grain weight of two mutants and ZH11 (n = 10). f Measurement of seed length, seed width, and seed thickness of ZH11 and flo19 mutant grains n = 10 1 3 of ZH11 from 7 DAF with the difference reached the maximum at 23 DAF, and then decreased slightly (Fig. 2d), which finally resulted in the decrease of 1000 grain weight (Fig. 2e). Both the two mutants showed reduced grain thickness, while only flo19-2 showed increased grain length (Fig. 2f). In addition, TTC staining assays showed that the seed viability of flo19-1 was significantly lower than that of the control, while flo19-2 was not affected (Fig. S3a, b).

Effect of FLO19 gene mutation on rice quality
The occurrence of floury endosperm phenotype is often closely related to the changes of storage substances in endosperm. Next, we evaluated the content of starch and protein, the two major components in grain endosperm. Starch consists of two members, amylose and amylopectin. Due to the lack of a reliable method to evaluate amylopectin content, only the amylose content was measured. Compared with ZH11, flo19-1 showed a similar value of amylose content, while flo19-2 displayed a significantly lower value (Fig. 3a). In order to reflect the change of starch more accurately, the contents of three sugars (sucrose, glucose, and fructose) were measured. Compared with ZH11, flo19-1 showed significantly lower values in all three sugars, while flo19-2 only had higher value in fructose (Fig. S4a-c). The majority of protein in rice endosperm is storage protein, including glutelin, prolamin, globulin, and albumin. Compared with ZH11, flo19-1 showed a significantly higher value in glutelin, but lower values in prolamin and albumin (Fig. S4d). In contrast, flo19-2 displayed lower values in prolamin, albumin, and globulin, but not in glutelin (Fig. S4d). The total amount of storage protein in flo19-1 seeds increased significantly while that in flo19-2 seeds remained basically unchanged (Fig. S4d).
To determine whether changes in starch and grain storage proteins accumulation were reflected by altered messenger RNA levels, we tested the expression of 81 key genes involved in grain storage materials . Compared with ZH11, the expression levels of 39 genes related to starch biosynthesis and metabolism in mutant flo19 changed: 16 genes were downregulated and 23 genes were upregulated in flo19-1, while 19 genes were downregulated and 20 genes were upregulated in flo19-2 (Fig. 3b). The expression levels of 17 genes related to storage protein biosynthesis and metabolism changed: 4 genes were downregulated and 13 genes were upregulated in both flo19-1 and flo19-2 (Fig. 3c).
The changes of storage substances in endosperm are likely to affect cooking and eating quality. Thus, we examined the gelatinization properties of milled rice flour. Rice flours from flo19-2 yielded a significantly better RVA curve pattern than those from ZH11 and flo19-1, reflected by a significantly higher breakdown value (Fig. 3d). Similarly, flo19-2 had a higher value than ZH11 and flo19-1 in gel consistency (Fig. 3e). In addition, there was no difference in gelatinization temperature between ZH11 and flo19 mutants (Fig. S4e).

Possible causes of phenotypic differences between mutants and phylogenetic tree analysis
The fact that editing the same gene produces rice lines with different cooking and eating qualities is an interesting point worthy of further exploration. We next examined temporal and spatial expression patterns of FLO19 using eight tissues from ZH11 and flo19 mutants (Fig. 4a). FLO19 was expressed constitutively in all examined tissues and showed the highest expression in the root (Fig. S5). The expression level of FLO19 was abnormally higher in the root and stem of flo19-1, and lower in other tissues than that of ZH11 (Fig. 4a). Similarly, higher expression level of FLO19 was only observed in the root and seedling of flo19-2 (Fig. 4a).
Comparative sequencing revealed that the allele from flo19-1 had a 1-bp deletion and that from flo19-2 had a 2-bp deletion (Fig. 1c), and these mutations directly led to varying degrees of shifting mutations and premature termination of the FLO19 coding products (Fig. S6). The normal FLO19 gene was predicted to encode a protein consisting of 293 amino acids with a GATase domain, a HTS domain, a Peptidase_C26 domain, and two unknown-function LCR domains (http:// smart. embl-heide lberg. de) (Fig. S7). It is worth noting that both of the mutated FLO19 proteins lost HTS domain and Peptidase_C26 domain, and the FLO19 protein of flo19-1 mutant even lost the LCR domain in the C-terminal (Fig. S7). Considering the phenotypic differences between the two mutants, we hypothesized that the LCR domain in the C-terminal may be critical for FLO19 protein function in CEQ regulation.
In order to further understand the phylogenetic relationship between FLO19-related proteins in plants and eukaryotes, we searched and compared the predicted protein sequences of 13 different phylogenetic organisms. Based on the phylogenetic analysis, FLO19 seems to be a green-plant-unique gene and rice FLO19 may be an evolutionary product of fungi FLO19 (Fig. S8). However, these genes have not been functionally identified even in the model plant Arabidopsis thaliana.

The effect of FLO19 on carbon and nitrogen metabolism
The FLO19 gene was predicted to encode a class I glutamine amidotransferase (GAT1) and showed the highest expression in the root (Fig. S5), implying that this gene may be involved in the nitrogen assimilation process of rice. Thus, we conducted a 2-week hydroponic nitrogen treatment with different nitrogen levels for flo19 mutants and ZH11 in the greenhouse. In contrast to ZH11, both mutants showed decreased shoot length accompanied by an increase in root length, and this change could not be rescued by increasing nitrogen supply levels ( Fig. 4b-d). Subsequently, we determined the activities of key nitrogen assimilation-related enzymes and glutamine amidotransferase for seedlings. Interestingly, just as the two mutants of flo19 differ in quality traits, the two mutants also differ in enzyme activities. Compared with ZH11, the enzyme activity of glutamine synthase (GS) in flo19-1 showed no difference in shoot under all nitrogen conditions except for 0 N nitrogen level, but was significantly reduced in root (Fig. 4e, f). In contrast, the activity of GS in flo19-2 was significantly lower in both shoot and root except for 2 N in the shoot (Fig. 4e, f). The glutamate synthase (GOGAT) activity of flo19-1 was significantly lower under all nitrogen conditions in both shoot and root, while it was just the opposite in flo19-2 (Fig. 4g, h). The asparagine synthetase (AS) activity of the two flo19 mutants was higher than that of ZH11 in shoot, with the exception of 0 N and 4 N nitrogen levels (Fig. S9a). Meanwhile, the AS activity of flo19-2 was only different from ZH11 under partial nitrogen conditions (0 N, 1 N, and 2 N) in the root, while no significant difference was observed in flo19-1 under all nitrogen conditions (Fig. S9b). Glutamine amidotransferase (GAT) activity result showed that flo19-1 had no difference with ZH11 under almost all nitrogen levels in the shoot except 0.5 N nitrogen level, while flo19-2 had lower values under almost all nitrogen levels except 2 N nitrogen condition (Fig. 4i). Meanwhile, flo19-1 showed no difference under almost all nitrogen conditions in the root except 1 N nitrogen level, and conversely, flo19-2 exhibited lower activity under almost all nitrogen conditions, except 4 N nitrogen level (Fig. 4j). The above results indicated that the nitrogen assimilation process in flo19-1 was obviously suppressed, while the inhibition of nitrogen assimilation in flo19-2 seemed to be improved to some extent.
Carbon and nitrogen metabolism is crucial for plant growth and development. Thus, we also determined the activity of Rubisco (ribulose bisphosphate carboxylase oxygenase), a key enzyme for CO 2 assimilation into the biosphere, in flo19 and ZH11 seedlings. Compared with the wildtype ZH11, Rubisco activities in these two flo19 mutants were significantly reduced (Fig. S10a), demonstrating that the FLO19 gene mutation caused damage to the carbon assimilation of rice. As an important metabolic intermediate, the concentration of acetyl-CoA not only reflects the general energy state of cells (Shurubor et al. 2017) but also affects the specificity and activity of a variety of enzymes (Pietrocola et al. 2015). The concentration of acetyl-CoA in flo19-1 seedlings varied with nitrogen supply levels, which seemed to be irregular, while flo19-2 showed significantly higher acetyl-CoA levels at all nitrogen levels (Fig. S10b). This indicated that the energy metabolism disorder in the flo19 mutants or the metabolic synthesis of organic matter was blocked; Fig. 3 Effect of FLO19 on quality traits. a Amylose content in milled rice flour (n = 10). b Expression levels of 39 genes involved in synthesis of storage starch in endosperm of flo19 mutant lines. c Expression levels of 17 genes involved in synthesis of storage protein in endosperm of flo19 mutant lines. d RVA profile. e Gel consistency (n = 10). Data are means ± SD (Student's t-tests, *P < 0.05; ns, no significant difference). In b and c, only genes with significant differences are shown here. Expression levels were determined by qRT-PCR using RNA samples from endosperm at 7 DAF with three biological replications ◂ Mol Breeding (2021) 41: 36 Page 9 of 15 36 thus, the normal growth and development were disturbed.

Haplotype analysis of FLO19
In order to investigate natural variations in FLO19, we analyzed variations in the coding region of FLO19 from the association population, and classified it into 11 haplotypes, of which Hap 1-4 were the main haplotypes (Fig. 5a, b and Data S1) and were used for further analysis. Hap 2 was mainly distributed in japonica rice subgroup, while the other three were primarily distributed in indica rice subgroup. Based on the phenotypic data of 533 varieties investigated previously Zhou et al. 2020) and recently (Data S1), we conducted a comprehensive analysis of quality and yield traits of these four haplotypes. Notably, Hap2 had the lowest chalkiness rate, amylose content, and gelatinization temperature, whereas no significant differences in chalkiness rate and gelatinization temperature were found between Hap2 and Hap4 (Fig. 5c, e). For protein content, there was no statistical difference among the four haplotypes (Fig. S11a). Hap2 had higher values in gel consistency and taste score than the remaining three haplotypes (Fig. 5f, g). Although Hap2 had the highest 1000 grain weight, its spikelet number, seed setting rate, and yield were similar to or lower than those of the other three haplotypes, and there was no difference in heading dates among the four haplotypes ( Fig. 5h and Fig. S11b-e).

Discussion
Rice endosperm is known to be an excellent system for elucidating how gene networks regulate starch synthesis and amyloplast development (Nelson and Pan 2003;Satoh and Omura 1981), so various mutants with defective endosperm have been screened for further research. In terms of appearance, floury endosperm mutants can generally be divided  Yu et al. 2020)) and partially floury endosperm mutant (such as FLO4 (Kang et al. 2005), FLO7 , and FLO15 (You et al. 2019)). Interestingly, similar to previously reported flo7 mutant, the peripheral region of flo19 mutant grain also appeared floury-white (Fig. 1e). Differently, a growth arrest of amyloplasts occurred in the flo7 mutant, but at least the shape of amyloplasts was intact, unlike the broken amyloplasts in flo19 mutant (Fig. 1h, i), suggesting that the damage caused by FLO19 gene mutation may be more serious. Notably, the expression level of FLO7 was the highest in the developing endosperm, while FLO19 was the highest in the root (Fig. S5), indicating that these two genes had different mechanisms of action affecting quality traits. The qRT-PCR results showed that among the detected genes related to starch and protein synthesis, these five genes (TPS, MAPK4, Wx, ABP, and DRP) had the most significant changes in expression (Fig. 3b  . Except for the Wx gene encoding granule-bound starch synthase (GBSS), which directly controls the synthesis of amylose, the remaining four are not directly involved in the synthesis of starch or protein. Although two mutants, flo19-1 and flo19-2, were produced by editing the same gene, the characterization results of various phenotypes were different. flo19-1 significantly increased storage protein content and reduced all three sugar contents with no difference observed in amylose content, while flo19-2 decreased amylose content and remarkably improved fructose content without difference in protein, sucrose, and glucose contents ( Fig. 3a and Fig. S4a-d).
An increase in fructose content and a decrease in amylose content of flo19-2 mutant may indicate that the conversion from sugar to starch is impeded. The decrease of three sugar contents in flo19-1 mutant may be related to the transformation between sugar and protein. These differences in biochemical parameters also led to the differences in gel consistency and gelatinization characteristics between the two mutants ( Fig. 3d, e).
The key nitrogen assimilation enzyme glutamate synthase (GOGAT) displayed significantly lower activity in flo19-1 than in ZH11, under all nitrogen concentrations in both shoots and roots, while it was just opposite in flo19-2 (Fig. 4g, h). The activity of glutamine synthase (GS), another key enzyme for nitrogen assimilation, was significantly inhibited in shoots and roots of both flo19 mutants (Fig. 4e, f). Although there was no statistically significant difference in the shoot of flo19-1 mutant, the values were lower than those of ZH11 at almost all nitrogen concentrations except 0.5 N nitrogen concentration (Fig. 4e). This may indicate that the main process of nitrogen assimilation in flo19-1 mutant was completely inhibited, while in flo19-2 mutant, although the enzyme activity of GS was also inhibited, the improved enzyme activity of GOGAT possibly alleviates the negative effect of suppressed nitrogen assimilation to some extent, which can be reflected in the differences in quality traits (Fig. 3a, d, e and Fig. S4), plant height (Fig. 2a, b), and seed vigor (Fig. S3a, b) between the two mutants.
In essence, the differences in editing sites result in completely different coding products of the FLO19 gene in these two mutants. Compared with ZH11, the FLO19 protein in the flo19-1 mutant terminated prematurely after changing five amino acids, while the FLO19 protein in the flo19-2 mutant continued to translate a fragment containing 35 aa from the mutation site (Fig. S6). It was this extra amino acid sequence that allowed the mutated FLO19 protein to regain an LCR domain (although it may not be exactly the same as the original) (Fig. S7), which we hypothesized might be responsible for the functional differences between the two proteins. But the significant decrease of GAT activity in flo19-2 seemed to argue that the newly generated LCR domain has caused more serious consequences (Fig. 4i, j); certainly, more assays are needed to prove it.
Indica usually has higher nitrogen uptake capacity and nitrogen use efficiency than japonica (Hu et al. 2015;Islam et al. 2021); thus, the yield per plant of japonica is often lower than that of indica. Haplotype analysis revealed that Hap2 could improve both appearance quality and CEQ of rice, but the yield per plant of Hap2 was also the lowest among the four main haplotypes (Fig. 5h). Considering that Hap2 mainly exists in japonica rice varieties, while the other three haplotypes mainly exist in indica rice varieties ( Fig. 5b and Data S1), the disadvantage of Hap2 in yield is more likely to be caused by the differentiation of indica-japonica subspecies than the FLO19 gene itself. In consequence, we advocate haplotype Hap2 as having great potential for rice quality improvement, especially for high-yield indica rice varieties.

Data availability
The variation information of 533 rice accessions can be obtained through the website RiceVarMap (http:// ricev armap. ncpgr. cn/). The phenotypic data of 533 rice accessions can be obtained from the references mentioned in the main text or the Supplementary Data part of this study. For materials, please contact the corresponding author's email address.

3
Declarations Ethics approval and consent to participate Not applicable.

Consent for publication Not applicable.
Competing interests The authors declare no competing interests.